Randomized controlled trial is the gold standard to evaluate the efficacy and safety of one treatment over another. While randomization ensures the comparability of treatment groups, randomized clinical trials usually provide average treatment effect as a summary result, which is less meaningful to physicians who treat individual patients. Alternatively, predictive approaches to treatment effect heterogeneity (PATH) are recommended to estimate heterogeneity of treatment effect across individuals. This method can predict risks of adverse events after treatment and thereby to stratify patients at a low to high risk for adverse events. In this thesis, we found that the SYNTAX score and its variants (the CABG SYNTAX score and residual SYNTAX score) could identify patients at higher risk of mortality after myocardial revascularization, who require more intensive pharmacological treatments for secondary prevention and need to be closely monitored during follow-up. Another strength of predictive models is to estimate which of two treatments will be preferred for individual patients when multiple patient characteristics that affect benefits or harms of treatments are taken into consideration simultaneously. In this thesis, we redeveloped the SYNTAX score II for the optimal mode of myocardial revascularization in patients with complex coronary artery disease, and we described the risks of a purely data-driven approach to discover a statistically significant predictive factor when developing models. The PATH can help the heart team provide patients and their families with a more transparent and shared decision-making process by providing objective, evidence-based information prior to treatments.